LLM Reference
GSM8KactiveMathematics

GSM8K: Grade School Math 8K

Metric: Accuracy (higher is better)Introduced: 2021

8,500 grade school math word problems requiring 2–8 steps of arithmetic reasoning. Widely used for evaluating chain-of-thought reasoning; top models now near 100%. High benchmark score alone doesn't make a model the right pick — weigh it against pricing, API availability, and release date.

Models ranked

10

tracked on this benchmark

Score band

95.7 – 80.6

best → lowest tracked

Snapshot trend

+2.37

May 28 → Jun 25 · 1 models

Leaderboard

Tracked models ranked by Accuracy (higher is better).

Compare candidates
#Model variant and provenanceScore
1
Llama 3.1 405B
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: May 28, 2026Confidence: Not recordedSource
95.7
2
Granite 4.1 30B
Version: 8-shot, instruct model (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
94.2
3
Llama 3 70B
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: May 28, 2026Confidence: Not recordedSource
93.0
4
Granite 4.1 8B
Version: 8-shot, instruct model (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
92.5
5
Nemotron-Labs TwoTower 30B-A3B Base
Version: 8-shot, accuracy; default TwoTower diffusion decoding at confidence_threshold=0.8, block_size=16, BF16 on 2xH100; evaluator/harness not published on the model cardHarness: Not recordedEvaluator: Not recordedObserved: Jun 25, 2026Confidence: Not recordedSource
90.1
6
Gemma 2 27B
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: May 28, 2026Confidence: Not recordedSource
89.7
7
Granite 4.1 3B
Version: 8-shot, instruct model (accuracy)Harness: Not recordedEvaluator: Not recordedObserved: Jun 7, 2026Confidence: Not recordedSource
86.9
8
Phi-3 Mini 128K
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: May 28, 2026Confidence: Not recordedSource
84.9
9
Qwen2-7B
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: May 28, 2026Confidence: Not recordedSource
82.7
10
Llama 3 8B Instruct
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: May 28, 2026Confidence: Not recordedSource
80.6

How to read this benchmark

This benchmark scores models where higher is better. Scores are useful for directional filtering and shortlisting — not for universal quality ranking. Prefer benchmarks closest to your workload, then validate the linked model pages for pricing, context window, and provider availability.

Trust this score when

  • There is a fresh timestamped snapshot (or multiple snapshots) for this benchmark.
  • The model list covers the same version family you can actually deploy today.
  • Top candidates overlap with your required routing and feature requirements.

Be cautious when

  • There is only one benchmark snapshot or the dataset appears stale.
  • The benchmark metric direction is opposite of your decision objective.
  • The score difference between options is narrow and likely within implementation variance.

FAQ

What does the GSM8K benchmark measure?

8,500 grade school math word problems requiring 2–8 steps of arithmetic reasoning. Widely used for evaluating chain-of-thought reasoning; top models now near 100%. On this page it lists 10 tracked model variants where higher is better.

Is a higher Grade School Math 8K score always better?

For this benchmark, higher is better. A high score helps you shortlist, but confirm pricing, context window, and provider availability on each model page before committing — the top scorer is not always the right pick for your workload or budget.

How current is this Grade School Math 8K data?

This benchmark was last reviewed on Apr 15, 2026. The tracked score average moved +2.37 points across the last 3 snapshots.

Related benchmarks

Last reviewed: Apr 15, 2026